I’ve transferred a discussion from Road Map to a separate thread here. Here’s a plot of monthly CET temperatures and a simple R script to download the data and make the plot. 
The warmest months are as follows: <><><><><
| JAN | 1916 |
| FEB | 1779 |
| MAR | 1957 |
| APR | 1865 |
| MAY | 1833 |
| JUN | 1846 |
| JUL | 2006 |
| AUG | 1995 |
| SEP | 1729 |
| OCT | 2001 |
| NOV | 1994 |
| DEC | 1934, 1974 |
I thought that AGW attribution and detection studes argued that winter months were supposed to be warming the most, but CET winter months are certainly not showing anything anomalous.
Climate Audit 
106 Comments
Re 124, Peter, you say:
Now that you mention it … yes, they do happen most every decade.
But that answer simplifies the situation immensely. In fact, the frequency of extreme events in natural records has long been known to follow a logarithmic decrease as the length of the record increases, which of course these climate events do.
Here are the number of record months, the expected (logarithmic) number of record months, and the average number of record months per year for England.
As you can see, the only slight deviation from the expected number is not happening now … it was back in 1993, when a few extra records were set. Since then, the count of extreme events (record months) has been about the expected number.
Thus, in answer to your question “is this record month unusual”, the answer is no. In fact, the last decade in England has not been unusual.
Thanks for the question, Peter, questions like that force me to take a fresh look at the data.
w.
nice chart.
I have to say Willis.
If it werent’ for the fact that you have a life I’m very jealous of, it’s obvious you could have suceeded well in any number of areas.
It’s not just the quality of your presentation, or the amazing distillarion of amorphous data, but the speed at which you can do it.
You could take over Wall Street in a few weeks of trading.
Re 144, thanks, great Viscous one. My motto was always “retire early … and often”. I think I mentioned that for the past three months I was working as a sport salmon fishing guide on the Kenai River in Alaska, fourteen hours a day, no internet at the resort, no time. Now, I’m retired again, living in Hawaii with my wife and daughter, and have time to enjoy my research … at least until the money runs out.
Not sure what’s next, for October I have an offer of a one month job in Oregon as a carpenter, in a place where I can fly ultralight aircraft with an instructor who is really good …
In any case, let me take this opportunity to thank you, and everyone, for the opportunity to participate here. I actually have a new plan for my place in all of this, which is to ignore ad hominems and personal attacks, to support everyone regardless of their sins of omission or commission, and just focus on getting the information out there in the world for people to judge and comment on. That’s the great game of science, put it out there and let people try to tear it apart.
What sparked it for me was Peter’s somewhat rhetorical question about whether a hot July was unusual … and what I realized was, I could respond to Peter’s tone, or I could respond to Peter’s question. What I saw was that at the end of the day, Peter and I are the same, we’re both, as the song says, ” just fools whose intentions are good.” So I decided to actually answer his question as best I could, to treat it as a real and serious question, and let all the rest go.
w.
Re #142, Willis, you misuderstand what I said. I know warmest months happen with a certain set frequency – even if the climate was cooling… presumably? That is why I said July was the warmest EVER month. How often do warmest EVER months come along?
Oh, and if you think the last decade or two of weather in the UK haven’t been unusual you must far away from here on an island in the Pacific or something…
Peter, Willis just showed you, in #142, how often warmest EVER months occur. They occur almost every decade. What more do you want?
You say the weather in the UK has been unusual in the last 10-20 years. Unusual compared to what? Your childhood recollections? I moved to Sydney about 20 years ago. I think that compared to then, there are fewer thunderstorms than the late 1980s. Whether that’s actually true, I don’t know, but even if it were, that only suggests that climate is variable.
As an amusing anecdote, last year I heard an interview with a climate scientist on Sydney radio, arguing that Sydney’s climate had changed due to AGW. The host invited listeners to ring in and say whether they thought the weather had changed. He fielded a volley of calls, all of whom agreed that the climate had changed in their lifetime, in that it used to be much hotter!
(Pure speculation, but I suspect this observation might be correct, in a perceptual sense. If you look at photos of Sydney from the 1940s and 1950s, it’s very clear that there were fewer suburban trees, and the trees were much smaller. These days, large parts of Sydney are very “leafy”, and this might contribute to perceptions of the temperature, given the great increase in the amount of shade. I also understand that some time ago Phil Jones changed his Sydney observation site from leafy Observatory Hill to the tarmac at Sydney Airport.)
James, I mean the warmest month – the Central England Temperature was aroung 19.8/9C. No month has ever been as warm. Ok, we’ve seen other warmest months (I remember how absurdly above average November ’94 was, or October last year) but, if we see a July or any month warmer than 06 (so, above 20C mean) soon then something remarkable is going on. July 06 was the warmest year in the Central England series going back to before 1700.
There is, though, another interesting fact. There have been slightly warmer 30 day period. In the amazing summer of 1976 to be precise (we’ve not seen anything like that since) and during 1995. Otoh, I think the persistence of general mildness these last few decades is highly unusual and not dismissable – it’s happening. Why? That’s more of a question imo.
Peter, I’m sorry, I don’t understand. You asked how often we see the warmest ever month, Willis showed that that occurs almost every decade. What do you want, a “warmest month” earlier than the 2006 “warmest month”? The warmest month is by definition the warmest month ever (at least in the instrumental record).
Your statement that “if we see a July or any month warmer than 06 (so, above 20C mean) soon then something remarkable is going on” is nonsense. There isn’t anything remarkable about a warmest month this decade. If there wasn’t a “warmest month” this decade (or next decade) that would be unusual, although not particularly interesting.
Why don’t you go and read #142 again?
For Willis, just for fun, why don’t you plot the “coldest month” record as you did in #142? Perhaps that would make the point for Peter?
PS, I have no idea what the “coldest record month” data looks like, but I would anticipate it looks much like the “warmest record month”. That seems like a fair test of whether the latest “warmest month” is exceptional. Willis?
James, sorry if I’m not being clear. July was the absolute warmest month in the last 300 years. It’s highly unlikely any other month of the year, other than perhaps August, could be as absolutely/relatively warm.
Wrt coldest records, try this from Ukww:
The length [how long since it happened] of warmest month CET records.
January record: 90 years
February record: 227 years
March record: 49 years
April record: 141 years
May record: 173 years
June record: 150 years
July record: 0
August record: 11 years
September record: 276 years
October record: 4 years
November record: 11 years
December record: 31 years (71 years)
The length of the coldest month CET records
January: 311 years
February: 59 years
March: 332 years
April: 169 years
May: 308 years
June: 331 years
July: 190 years
August: 96 years
September: 199 years
October: 266 years
November: 226 years
December: 116 years
Right, I’ll shut up else TCO will be really, really bored…
Peter,
Peter, you are not being clear. I’m not disputing that July 2006 was the warmest in the record. What we are discussing is the frequency of record events. If Willis is kindly disposed, I’d like to see the frequency distribution of “coldest months”.
Your figures do not say anything about that distribution. All they suggest is some recent warming, which is not in dispute (from me anyway). What we are talking about is the significance, or otherwise, of a “record month”.
I believe that this is Willis’ point, and I note, from your data that the warmest September was 276 years ago. “Record” temperatures are not a particularly useful metric.
PS I just noticed that Willis in #142 only mentions “record” months, not record “warmest” months, so we need some clarification.
Switching from Central England temperatures to global-mean temperatures. In evidence before the Select Committee on Economic Affairs of the UK House of Lords on 18 January 2005, Sir John Houghton, co-chair of the IPCC Working Group that produced “Climate Change 2001: The Scientific Basis” said:
“…[S]ince 1990 we have had a continuous increase in global temperatures, a steady, consistent increase. The year 1998 happens to be the warmest year on record, and a more striking statistic than that is that each of the first eight months of 1998 was the warmest month on record of that kind. Now after you get eight ‘ducks’ in a row it is trying to tell you something…”
The GISS monthly average global surface temperature for August 2006 has just been published, and is lower than that of August 1998. The same is true of six of the seven previous months in 2006, compared with the corresponding months in 1998 (the exception was higher than in 1998 by 0.01 C).
NONE of the first eight months of 2006 was a record for the month – rather surprising for a series which is supposed to be rising continuously, steadily and consistently. Maybe it’s trying to tell you something.
Re #11
I think its surprising that John Houghton is still trying to sell the Hockey Stick’s conclusions as if they were still credible.
Willis’s point (that for a gaussian distribution extension of the record will give more records at some rate) is valid. But it is also valid that we are in a period of warming and that the rate of records is higher than it would be purely based on guassian statistics. There have been papers written on this. Those papers have been mentioned here and at RC. Next case. [bangs gavel]
Switching back to the Central England record, I now realise that Peter is making the point that July 2006 is not just a record month, it is, in absolute terms, the highest monthly average EVER (meaning any month).
However, if July is generally the warmest month in the UK (I’m not saying it is), then that contributes nothing to the discussion of the frequency of record events. If July is the warmest month, then to say that a record July is the warmest month EVER (in an absolute sense) is not surprising. In fact it’s obvious. The question is whether it’s unusual.
#13. TCO – why would you say that temperature statistics are purely gaussian? Better unbang that gavel. Rasmus had an idiotic post on that which I discussed last year. First they are autocorrelated. Record setting events occur under very different circumstances under such cases – with long-term persistence, they tend to occur in bunches. Also the distributions tend to be long-tailed as well as autocorrelated. Mandelbrot talked about this years ago. This should be in the repertoire of climate scientists talking about climate statistics. Willis’ plot is prior to consideration of any such issues which change things a lot.
Re #14, yup, it’s the warmest in at least 300+ years. So, not very unusual…
Just wanted to comment the long-term average July temperature in the UK in actually -5C.
For 115,000 years at a time, the average temperature in July is about -7C, on top of hundreds of metres of ice that is. Then there are periods of 15,000 years when the average July temperature is 19C.
Re #17, well thanks Jeff for putting it into context ;). Now, I’m off to try and thaw out the plumbing, it must -5C out there like you say? Well, it’s September so it must be colder than that?
This is now. Apples aren’t oranges…
Steve,
1. I’m clarifying the difference in mindsets: records will be broken at some frequency purely by chance. However, if they are being broken faster then normal, this may indicate something is up. Neither getting flustered about an individual record being broken nor dismissing record breaking tendancy are appropriate. Your point is at a further level of sophistication.
2. Did Willis use some model which accounted for autocorrelation, hmm?
3. ARe you sure that the rise in 20th century temps is “autocorrelation” or could it be CO2 induced warming. How can you resolve the difference?
re #19. TCO, point 1 taken. Points 2 and 3, good questions.
This thread has wandered off to an island off Europe. The proper subject of this thread is the failure in 2006 to exceed the 1936 record temperature in the US, a much more consequential piece of real estate. Which might stand for some time, as I have just found this excellent paper on solar cycle 24:
http://www.leif.org/research/Polar%20Fields%20and%20Cycle%2024.pdf
The predicted length of this solar cycle suggests a temperature decline of 0.8º C to 2013, which will erase all the gain of the 20th century.
And words like “ever” and “normal” are just fine when speaking of the earth’s overall climate history. Riiight.
What happens to the record breaking temps if you take out the data for large cities in the UK which I assume have been growing, not shrinking in the last few decades?
Can someone identify a url for me giving monthly US temperatures in a digital database? I’ve spent an hour trying to forage through GISS and UHCn without any success although I suspect that it’s somewhere.
GISS is available from a Junkscience link.
#24. I’ve seen GISS annual data. I haven’t had trouble finding GISS or anything like that. I’m looking for monthly U.S. data. I’ve foraged through this data without success and I’m hoping for an exact link (not just a general website) containing monthly anomalies for the US.
RE#4
No I didn’t know that. I assumed you were still in Tahiti or some such, and they were’t going to get you out of there without an especially large crowbar.
Certainly sounds like your enjoying yourself.
I retired myself a few year ago for a few months. Was rather enjoyable, though I went nowhhere exotic. Made more money than ever in my life.
Realized I was retired when I went to Ireland for Christmas. For some reason at the rental car place they asked me for my occupation. I was stumped for a moment and the girl behind the counter was looking at me oddly. I explained to here that I had been laid off and hadn’t spent the time to get a new job yet. “So we’ll put you down a retired then”.
Was an enjoyable time. I reckon the weather in Hawaii is better though.
#13,15
Why Gaussian? Check Gut’s paper (and comment 143 here), the assumptions for log(n) record times are i.i.d. and absolutely continuous random variables. No Gaussian assumption.
#@7. OK i.i.d. suffices without requiring gaussian- but as soon as you abandon independence, it changes fast. Mandelbrot talks about exactly this in the context of persistent processes and observed that records occur in bunches.
#23: I don’t know if these qualify:
http://www.ncdc.noaa.gov/oa/climate/research/ushcn/ushcn.html
http://cdiac.esd.ornl.gov/epubs/ndp/ushcn/monthly.html
#29. I’ve looked through the first site and didn’t find it, but it could be there somewhere. THe second site, which I also checked, lists the information for 1221 stations which is interesting and useful info but not a monthly U.S. average temperature anomaly. I presume that there would be such a thing to yield the annual US anomaly which is available.
[I’ve transferred this from “Road Map” …]
Peter, in #146 (in Road Map) you point out correctly that I misunderstood your question. You ask:
How often do warmest EVER months come along?
Unfortunately, this question can’t be answered. The problem is the shortness of our datasets, because we don’t have temperature records back to EVER.
According the the Greenland ice cores, for example, we haven’t had a warmest month EVER for about 8,000 years, and this July didn’t even come close to the record set in the “Holocene Optimum”. We’re currently at the cold end of the Holocene.
According to the CET, on the other hand, the last time there was a “warmest ever” month was in 1984. July this year was 19.7°C, July 1984 was 19.5°C. It’s not clear what this means, however, for two reasons.
The first reason is that the CET has only been partially adjusted for the Urban Heat Island problem. Thus, we don’t know if this latest “record” is really a record or not. See here for a discussion of this difficulty.
(As an aside, the UHI was first noticed by an Englishman, Luke Howard. In The Climate of London (way back in 1820), he commented “Night is 3.70°[F] warmer and day 0.34°[F] cooler in the city than in the country.” Despite his finding UHI in 1820, the CET is adjusted only about 0.2°C for the UHI, and in a stepwise fashion, for the post-1960 period.)
Second, the confidence interval for the monthly CET is quite wide, at +/- 0.6°C (3 standard deviations). This means that two monthly temperatures have to differ by more than 1.2°C before we can say that one is significantly larger than the other. (We could use 2 standard deviations as our criteria, but with 4,172 monthly data points in the CET, we’d have over 200 erroneous conclusions using 2 SD, so we should adopt a stricter standard).
So with a difference of only 0.2°C between now and 1984, we don’t even know if the most recent July is warmer than the 1984 record or not. There is a full review of these uncertainties here.
So, after all that, the short answer to your question? The answer is, because of the shortness and uncertainty of our records, your question is unanswerable as asked.
w.
If you smooth the CET data first, recent decades stand out as being anomalous, with more decadal mean records being broken than expected. Because the data are autocorrelated (especially after smoothing) the usual expectation of logarithmic decline in record frequency cannot be used. Confidence intervals can be calculated by permuting the data. To allow for intra-annual dependencies, I permuted all months together, not independently.
urlnr]nr]-nr
records(smo.cet[newindex,])
})
ci
Re #31
Willis, I think you are still missing Peter’s point. I think I know what he is trying to get at. If I am right, Peter’s point is a bit moot really, but let me try to disambiguate his sentence (as how I read it).
What he means is that the last month (July 2006) is the warmest month on the CET record, that is to say, no other single monthly measurement (Jan thru to Dec) for any other year (16xx? thru to present day) has a warmer monthly measurement.
The point is moot because comparing July temperatures to January temperatures is… well.. absurd. Of course a record July is highly likely to be warmer than January. From an information theory perspective, Peter’s point adds little new or useful information.
What might be more interesting would be to calculate seasonal corrections for each of the twelve months, apply these to the records, and see if July 2006 is the warmest record. (Would it be necessary to normalise first and second order moments? Or just first?) This doesn’t add much to be honest – you’re still comparing apples and oranges – but at least it is a test with a level playing field.
Further to #31 there is a second very similar document reviewing the uncertainties in the CET at:
Click to access HCTN_50.pdf
-With a different take on the strange Rothamsted Tmin anomalies.
Re 33, Spence, your disambiguated question is what I answered in #31 … I think. The previous “warmest ever” month was July 1984, with the caveats I indicated.
Regarding seasonal corrections, it’s an interesting question. I tried it your way and found that the variance is much greater in the winter. Thus, the records are not set in the summer but in the winter.
By that measure, the most recent record in the CET was set in … wait for it … January 1916.
w.
If you center each monthly series over its entire length and divide by its standard deviation over its entire length (none of this 30-year short-segment standrardization crap), the warmest record is June 1846. July 2006 is the 5th warmest month in the league table on this basis. Again its my understanding that winter months are supposed to be the most influenced by additional CO2 and we don’t see records being set there.
#32. what smoothing did you use to make cet.smo?
Re #35, #36
Yes, I see you have addressed that point now re-reading it… slightly different to my way of addressing Peter’s point, but similar results.
Thanks for doing the calcs, I seem to be getting lazier in my old age, I think that is a fairer way to rank “warmest months on record across all months”. It seems July 2006 doesn’t quite hit the highs with this method.
Willis, you may get some useful mileage out of Numberwatch. John Brignell did quite a bit on the statistics of extreme events, though whether he managed to pair that up with autocorrelation is beyond me…
cheers
per
Re 36, Steve M, standardization of the datasets is indeed the correct way to compare the extreme values. However, the lack of record high temperatures in the winter months does not correlate with the trend of each month. Here are the trends and their sigmas:
Note that the summer months do not have a significant trend, while the winter months do, and that the trends are larger in the winter months.
w.
PS: the CET record is unusual, in that the individual months are not very autocorrelated. The lag-1 autocorrelation of the twelve months ranges from a high of 0.18 (August) to a low of -0.05 (June) … why? Not a clue.
I suspect because we have very “interesting” weather, hot and dry with a continental influence one day can turn into cool and wet with an Atlantic flow the next and then back again a few days later.
Snow in Salt Lake City
Last winter was the coldest for 40 years in Europe.Spring had 2 false starts causing our (and our nieghbours) flowering shrub Lavateria to die.The late spring was followed by a warm July which caused bumper crop of blackberries at least one month earlier than usual.
At the end of July when the BBC was trying persuade the weather forecasters to say that August was going to be just as hot I forecast that it would be cooler and wetter than average. August duly obliged.
I am now forecasting that next winter will be cooler than last which should make 2006 the coolest in a decade.
Temp records according to raw proxies (data from Osborn Briffa 2006):
Warmest years:
1980
1434
1969
1517
1936
911
1989
1921
1831
1965
1691
1150
825
1989
Coldest years:
809
1699
1705
1947
1522
848
1491
1337
804
1818
1347
1237
1872
1648
And here is the big picture.
“Last winter was the coldest for 40 years in Europe.” since 1966 then? I’d like to see the figures, for the winter (which is December to February) that is. More like the very cold spell seen in January is being spun out of control and now it’s a ‘coldest in 40 years’ myth. Or perhaps you’re thinking of a forecast that was put out? Lets get this straight so this goes no further – agreed Tom? So, lets see the figures.
Certainly winter 05/06 wasn’t even cold across the UK.
What Peter. You mean like saying that a hot spell in July is the being spun as the “Hottest Summer” in 40 years myth?
You going to straighten that out so it goes no further Peter.
Your being a bit obvious Peter.
Defend the indefensible by attack eh 😉
“You mean like saying that a hot spell in July is the being spun as the “Hottest Summer” in 40 years myth?” Not by me, and not by anyone who knows July isn’t the UK summer (which is just about everyone who knows anything about meteorology…). So who has described it as you claim?
But, looks like I’ll have to repeat it again. July was the hottest July ever recorded (very close to ’83 in absolute terms though). June was warm, August marginally above average. Summer (which, to repeat, isn’t just July, but June and August as well) ’06 was a warm summer. One of several recently. I honestly haven’t seen how warm yet, or can find said, but certainly it WAS NOT hotter than ’76 and I’d guess ’95. OK?
Now, over to you Tom.
Peter you were with everyone else back in July when they we’re saying we were on the road to being the warmest year ever, just like you were on the hurricane season. Jut as you will be the next prediction that is in line with your agenda.
RE: #15 – It would appear that in July we had the odd circumstance of a classic “stagnant Atlantic basin” scenario (which typically results in warmth in both the East Coast of the US and Western Europe), at the same time as a persistent Gulf outbreak in the Mississippi Valley at the same time as a class AUTUMNAL triple barrel High with offshore winds out West. Talk about constructive interference. No doubt, persistence would have been in play.
RE: #42 – Normally that scenario does not occur until mid to late October.
Warm Summers in England Per the CET
Here’s the top dozen:
1976 _____ 17.8
1826 _____ 17.6
1995 _____ 17.4
2003 _____ 17.3
2006 _____ 17.2
1846 _____ 17.1
1983 _____ 17.1
1947 _____ 17.0
1933 _____ 17.0
1911 _____ 17.0
1781 _____ 17.0
1899 _____ 16.9
Again, not much sign of warming, second warmest summer was in 1826. This of course leads to the "warmest summer in 180 years, so it must be AGW" fallacy. Why is this a fallacy? Because it is likely that humans were not the cause of the 1826 high temperature, so there is no a priori reason to think that humans are responsible for this high temperature … Bear in mind that the error (1 std dev) of the monthly data is +/- 0.2°C. This means that the error of the three month average (1 std dev) is sqrt(3 * 0.2^2) = 0.35.
Even with a very generous requirement of p less than 0.5, none of the top 12 summers are statistically distinguishable. This is one of the problems of climatology, that we end up discussing and disagreeing about numbers that statistically are all the same … how many angels can dance on the the head of a a pin, anyhows.
w.
Willis, you might try using the tex feature.
Have a look at the rural Armagh Observatory Teperature Series:
Mean Annual, maximum/minimum temperature:
http://climate.arm.ac.uk/calibrated/airtemp/tm-an-maxmin-dtr.dat
Mean Seaonal Max/Min Temp:
http://climate.arm.ac.uk/calibrated/airtemp/tm-seas-maxmin-dtr.dat
Mean monthly max temp:
http://climate.arm.ac.uk/calibrated/airtemp/TMAXC-MON-C.DAT
Mean monthly min temp:
http://climate.arm.ac.uk/calibrated/airtemp/TMINC-MON-C.DAT
Plot of Armagh v CET Annual means 1844-2004:
http://www.junkscience.com/MSU_Temps/Armagh_vs_CET.htm
Re #51, Willis, shouldn’t you be dividing by 3 not multiplying by 3.
The Central England Temperature data between 1707 and 1722 were observed in The Netherlands. As no data existed in Britain Manley used the Labrijn series and adapted it to fit the other data. (Reminds me of Arrhenius somehow).
Here is the complete Labrijn series corrected for clerical errors By van Engelen and Nellestijn. When using it please quote this source:
Here is the 30 year moving average for every month (with theoretical CO2 forcing on top):
Here is a comparison with the definition of climate being dependent the average of the coldest and the warmest month. Holland had always a Cfa climate even in the LIA.
Note also that LIA is a winter phenomenon.
correction: of course a Cfb climate
Re 51.
Don’t be so fast to reject AGW with these numbers. Three of the warmest dozen summers in the CET record (since 1659) are in the last 20 years. What is the probability of this happening by chance?
I cannot remember the code to find this probablity exactly, so the following code finds this by simulation:
cumsum(rev(table(replicate(10000,sum(sample(2006-1659,12)The probability of having three or more summers from the last 20 years in the warmest 12 is ~0.025. This is significant at the p=.05 level.
Of course, this is a post hoc test. An a priori test would be much more persuasive, but we would have to wait 20 years to do it. Would it be wise to leave any control on greenhouse gases till then?
The code appears to have been garbled, here it is again
x
I don’t think that anyone denies that the end of the 20th century is warm within a 400 year perspective. In htat context, it’s surprising that the CET record isn’t louder.
steve if you plot the data as points instead of lines you see interesting sample artifacts in the early part of the dataset, these were the non-fahrenheit-quality thermometers.
Re 57, Richard, thanks for your post. You say:
This is a common, but fallacious, argument for AGW. In fact, it is what we would expect from a warming earth. We know that the earth has been warming since the Little Ice Age, so all that the data shows is that yes … the earth is warming. It says nothing about whether the warming is anthropogenic.
Finally, please re-read my comments on the CET and the UHI, as well as size of the standard error of uncertainty in the CET. In addition, the CET is not a homogeneous record. It is a composite of a varying selection of stations, including some data (as Hans Erren points out above in #55) from the Netherlands. In particular, the method used for calculating the monthly CET was changed in 1930, 1974, and 2004. Also, the method up until 1930 involved using between four and seven stations, depending on availability of the data.
Because of these changes, the CET is nowhere near as reliable as a single station record regarding long term trends. Here, for example, is a very long nearby single station record, that of Armagh Observatory in Ireland:
Note that, unlike the CET, this record shows absolutely no sign of any modern acceleration of warming. This is typical of many rural stations world wide, whereas many urban stations generally show increasing warming, presumably from UHI.
w.
#57 — Statistical probability is not physical probability. It’s too easy to forget that in the climate proxy business. Low statistical probabilities don’t tell us squat about whether some on-going temperature trend or some recent temperature anomaly is physically unlikely.
#40: Great post. You are really making good points lately. Maybe I judged you too harshly for the division of accuracy over time misconception. Maybe you are not a silly…like jae or such. Seriously, good job, man.
Re 63, thanks, TCO. I do my best …
w.
Heikki Tuomenvirta’s recent PhD thesis Reliable estimation of climatic variations in Finland might make interesting reading for those interested in instrumental records. Although the scope of the thesis is geographically rather limited 😉 , I’d like to point out two things which might be of a more general interest:
1) Although the author correctly notices that the CRU temperature record (Jones&Moberg, 2003) is unbiased with respect to the “true” Finnish temperature during the 20th century, he fails to notice that the difference is not equally distributed along the time span, see figures 5.2 and 5.3. Especially, the CRU record underestimates by about 0.2C the coldness in 1900’s and the warmness in the mid 1930’s. This has an impact on the relative variations during the century: in the “true” series the temperature raises about 1.5C from 1900 to late 1930’s, then cools, and finally raises about 1C from the late 70’s to the current temperatures around the same level as those in 1930’s. In the CRU version, the raise in the beginning of the century is only 1C and the current temperatures exceed those of 1930’s by 0.2-0.3C. It would be interesting to know if the difference in CRU record is something specific to Finland (or if there is a reasonable explenation for the difference, e.g., due the larger spatial averaging used in CRU) or if for some reason CRU record is biased also in other places in the beginning of the 20th century. Anyone know of similar studies (comparisons with the CRU record) from other places?
2) The author finds no precipitation trends at all. This is interesting as all climate models are predicting increased precipitation to Finland, see FINSKEN (where Tuomenvirta is one of the principal contributors).
As a side note for those regulars in the blog seeing bias in every study linked here: Tuomenvirta is an outspoken pro-AGW scientist (at Finnish Meteorological Institute) regularly appearing in Finnish newspapers. So definitely no anti-AGW bias in his study.
#65
1930’s are warmer than present ‘almost everywhere’. Even average of the proxies show this.
re #66: Yes, I’m aware that many instrumental series (especially in NH) show equal/higher temperatures in 1930’s. What I’m interested is that if there are actually other studies suggesting that there might be something wrong in CRU representation of the beginning of the 20th century. It may be as well that there are areas where the 30’s was cold or that “CRU bias” is of an opposite sign, and therefore the CRU representation is correct. As you well know, the temperature history accordind to CRU is this:
Hmmm, the image did not show up.
re: 61
I intensively looked at UHI in Europe. I compared Metropolitan Brussels with urbanising De Bilt and verified the correction with rural Fichtelberg.
http://home.casema.nl/errenwijlens/co2/homogen.htm
I then looked for other long series in Europe notably Hohenpeissenberg on a rural hilltop in Germany and graphed the data together:
This is guaranteed classic UHI free data. I don’t know about landuse changes that have effect on rural areas but that isn’t classic UHI anyway. Note that The Hohe Warte Station in Central Vienna has no observable UHI change as Vienna was already a Metropolitan area in 1780.
The match with Jones is surprisingly good:
giving the fact that he totally screwed up Brussels:
see also:
http://home.casema.nl/errenwijlens/co2/europe.htm
So think carefully before you shout: “UHI!”.
#67, Jean S
What possible reason would someone from a fine upstanding UK educational insitution have to ‘bias’ their temperature record data so that it did not show temperatures in the 1930s as high as they are now? Could it possibly have anything to do with the fact that it might undermine the case for AGW as it would mean that the earth actually warmed significantly during the earlier half of the 20th century when CO2 levels were somewhat lower than they were post the 1940s – god forbid it might even provide some evidence that increases in CO2 could actually be following on from increases in temperature rather than the (AGW) other way round?
Here is a relevant link which might explain the need for the ‘bias’. So it looks like we’ve only got four years left in the UK to build all those wonderfully environmentally (zero impact) friendly wind turbines and short down all our fossil fuel electricity generation plant.
KevinUK
Human metabolic “forcing?”
A human emits about as much energy as a 110 watt light bulb (120 for men; 100 for women, according to Wikipedia). Therefore, it can be easily calculated that human metabolism is causing a global temperature “forcing” of 1.2 watts/m^2. Looking only at land surface area, where all the people are, it comes out to 3.9 watts/m^2 (same as CO2!). Looking at very densly populated areas, it’s probably far higher than this. Wonder how much GW is contributed by this AGW?
(Assumptions: Earth surface = 5.096 x 10^11; population = 6 x 10^6; land surface area = 30% of total surface area.
Nuts: Forget it. divide everything by 1000. 1KM^2 = 1 million m^2, not 1,000 like I assumed.
Re 69, Hans, I’m not clear what you are saying. When I compare Fichtelburg with Brussels, the UHI of Brussels is quite evident. Are you saying there is UHI in the GISS records, or not, or what?
w.
Willis, I think Hans’ point is that these UHI free records still show the current warm peek to be above that from the 1930s and 1940s. Whether that’s true globally I have no idea.
Re 74, Dave, thanks, but if that’s what he meant … he sure didn’t say it.
w.
re 73:
When you compare rural Hohenpeissenberg with the jones grid average for central europe (Jones grid cell 0E-15E 45N-55N) it is the same. Hence Jones for central Europe does not show a UHI. My version of Brussels (Uccle) – which is corrected for UHI using population statistics – is similar to rural Hohenpeissenberg and Jones gridaverage. Therefore the jones grid average for central europe is UHI free.
Which makes me very curious about the averaging method.
Of course that is absolutely no guarantee that Jones is worldwide UHI free.
Of course there is UHI is the GISS records. UHI corrections by GISS is a joke because the don’t take populaton change into account but simply use automated trend subtraction which makes the De Bilt case very funny. Here is a comparison of me (using population statistics and homogeneity adjustment) Van Engelen (without UHI correction but with homogeneity adjustmen)t and GISS. You see that van Engelen and I are close with our independent homogeneity adjustment and that GISS is the odd one out.
The graph shows the correction values which are subtracted from the raw data. Note that subreracts a cooling trend threby warming De Bilt after 1875.
The GISS homogeneity algorithm can’t handle jumps in the data.
NB The early data in van Engelen (1706-1735) are the observations by Nicolaas Cruquius which are used in an uncorrected form by Manley in CET.
garbled:
Note that GISS subtracts a cooling trend thereby warming De Bilt after 1875.
See here regarding a very topical new paper on CET trends.
‘In Brighton yesterday, Dr Stott said: “This is a remarkable anthropogenic signal. Sharp spikes in warming have been recorded in regions across the world, but because we in the UK hold this unique temperature record stretching back nearly 350 years we are able to say that background climate ‘noise’ can’t reasonably be held responsible for what’s happening in Central England.
‘”This is the first time anywhere in the world that climate scientists have been able to look at a small geographical area, identify significant warming and say humans have very likely played a part.”‘
BBC say 1C from the 50’s
http://news.bbc.co.uk/1/hi/sci/tech/5357606.stm
Cherrypick a decade and make assumptions.
Hans, thank you for your information, which is very interesting, particularly the GISS correction for urban cooling!?! in De Bilt.
However, I’ve taken a look at the data for Jones Europe (45-55N, 0-15E), and compared it to three rural stations, Hohenhisenberg, Zugspitze, and Fichtelberg. While the post 1963 warming is the same for Hohenhisenberg and Jones Europe at 0.38°/decade, the Zugspitze and Fichtelberg records show only about two thirds of that warming, at 24°/decade.
Also, the GHCN data, which arguably does more to adjust for UHI, comes in between those two extremes, at about 0.34°/decade.
Finally, it’s worth bearing in mind the error bars on the HadCRUT (Jones) data, which are on the order of +/-0.1° (95% confidence) for the annual data. Because of that, the 1963-2005 trend may be anywhere from about 0.32°C to 0.42°C … which makes it hard to determine whether there’s UHI or not.
All the best, thanks for the good information,
w.
Re #80: The analysis went back to 1700.
“The analysis went back to 1700.”
Which was in the coldest part of the LIA. Temperatures peaked in 1999 at an annual average of 10.63C. 2005 stands at 10.44C, a fall of 0.19C. The potential problems with the CET have been highlighted above.
Have temperatures peaked? Will they go on rising, or are they heading for a fall regardless of atmospheric CO2 concentrations?
The heatwave of 2003 is blamed for 2000 excess deaths in the UK. No-one seems to mention the 25,000 to 45,000 excess deaths each winter.
re #79: Everything said in your article could be said about the temperature record I linked in #65: Finland has warmed MORE than 1C since late 70’s (or since Beatles), she warmed more during the 20th century (0.8C) than the globe on average, and she has warmed almost 2C since 1870s. But still, where is the local AGW-signal (see figure 5.8)? I guess you need a well-tuned climate model to get it out… which Tuomenvirta also did. So SteveB, what’s actually new in the study? Did they also check the precipitation? Why not or why they did not publish it? Maybe the results were not fit to the scientific meeting at
SteveB, your political agenda is getting far too obvious.
My mistake, Figure 5.3 (or Figure 5.4).
#79 and #80, Steve B and Paul B
Are we really expected to swollow this BS from ‘Stottie’ and his old mate Karoly? Do they honestly expect us (in the UK) to believe their estimate of the probability of 95% (100 – 5) that the so called observed increase in average CET (based on an average taken from THREE weather stations!!) cannot be explained by natural variability (forcings) because a bloody Hadley GCM says so? Pull the other one it’s got bells on.
The real explanation and timing behind this BS is that it’s party conference time in the UK at the moment and this BS was put into the public domain at a ‘climate clinic’ organised by “Britain’s major environmental group – with The Independent as media partner” at the Lib Dems conference in Brighton this last week. The Independent has had global warming scare stories on its front cover (Arctic ice melting) and inside pages all this last week. Now I wonder who that “Britain’s major environmental group” is?
KevinUK
Is there a formal definition of background climate “noise’ ? If yes, what is it?
Re #84: Pay attention, Jean. They’re doing these clinics at each of the party conferences. Recall also which UK party now has the most aggressive platform on AGW. That turn of events must make KevinUK very sad.
re #88: Sure, and doing clinics at each of the party conferences makes these clinics scientific and objective without agendas. Geez, don’t expect me to take you seriously even if you got invited to all of the national conventions in 2008. SB, you are the first person to get into my “ignore completely”-list. Congratulations!
Isn’t there a rule against discussing politics here?
Dear friends, the CET can’t prove anything. Why? Because it’s not a continuous record.
From 1878 to 1930, the monthly CET mean temperature record was the average of “Lancashire” and Oxford, where “Lancashire” was derived from four to seven stations in the northwest of England reduced to a common standard (Manley, 1946)
From 1931 to 1973, the Oxford record was thrown out and replaced by the “corrected” Radcliffe Observatory monthly mean (Knox-Shaw and Balk, 1932).
From 1974 to October 2004, the whole lot was thrown out and replaced by Rothamsted, Malvern, and the average of Squires Gate and Ringway.
And since October 2004, it’s been comprised of Rothamsted, Malvern, and Stonyhurst, equally weighted.
Now, any scientist worth his salt will tell you that while this makes for an interesting record, we cannot draw any kind of firm conclusions regarding the record. In particular, comparing the pre- and post- 1974 records is useless.
Having said all that, here’s the record so you can judge for yourselves whether the current warming is unusual …
Some items of note. First, the recent warming is not anomalous. The current 25-year trend is at a rate of about 5°C/century, and this rate has been met or exceeded (both in terms of warming and cooling) a number of times in the past.
Second, none of the temperatures stray very far from a linear trend, including the most modern section.
Third, although it is hard to see at this scale, the 25 year trend has been steady for the past 6 years.
But, like I said at the start, it’s all moot because THIS IS A SPLICED RECORD involving a changing cast of weather recording stations, so no firm conclusions can be drawn.
w.
RE: #91 – That is but ugly. What a mish mash of sites!
re 81:
do you have a graph? It’s nice bringing in new data to the argument but I would then like to see some quality asessment first. The data I am displaying for europe show a coherent picture.
Also displaying a graphs with a 35 degrees y-axis range is covering up salient properties of the dataset. Please show winter temperature and summer temperature seperately.
Forgive if I’m making a foolish error or misunderstaning something, but (assuming for each specific month the random variables are independent and identically distributed between years) I believe that after y years, the expected number of record months in a year wouldn’t be related to log(y), but would just be 12/y (12, because of 12 months).
Consider July after y years. The probabilty of the record occurring in any particular July is 1/y. So, the probability the record occurred in the latest July is also 1/y. Same for all the other months. So the number of records in the year is just a binomial distribution with n = 12 and p = 1/y, which has an expected value of 12/y. (Of course, after 13 years the expected value would be less than 1, and the probability of no records would be [1 – 1/y]^12).
(I glanced at the Gut paper, but I don’t think it’s talking about the same thing. It seems to be talking about the total number of records set after n trials. That would be approximately logarithmic, since it would be proportional to the sum of integer recipocals.)
My university recently produced a 500 year Northern Hemisphere temperature reconstruction from speleothems showing an overall warming trend of 0.65K:
http://www3.interscience.wiley.com/cgi-bin/abstract/112604852/ABSTRACT
Looking at figure 1:
Compared to 1961-1990, the anomaly was -0.6C in 1550, about -0.35 by 1620, about -0.15 by 1770, -0.4 by 1850, -0.2 by 1880, about -0.3 by 1920, just under zero by 1950, about +0.05 by 1970, falling to about -0.75 around 1980. The instrumental then record takes it to +0.2 by 1990.
“The speleothem temperature reconstruction closely resembles that of Briffa et al. (2002) for the time interval ~1700 to the present. The overall magnitude of warming in the speleothem reconstruction is 0.65 K, compared with around 0.4 K displayed by the reconstruction of Briffa et al. (2002). The twentieth century, as with all the other reconstructions, is significantly warmer than the previous four centuries. The temperature is shown to increase by 0.35 K over the last century, compared with the 0.29 K increase that occurred between 1700 and 1780, and the 0.27 K amplitude change in temperature that occurred during the late sixteenth to early seventeenth centuries. With such a short dataset, it is difficult to assess whether the amplitude of the temperature increase during the twentieth century is truly anomalous relative to what occurred in the preceding centuries, but the recent warming does not appear as exceptional as that suggested by the borehole reconstruction of Huang et al. (2000), where the nineteenth and twentieth centuries are responsible for 80% of the net temperature increase over the last 500 years. The discrepancy between borehole and speleothem records may be due to the European weighting of the speleothem reconstruction, as the borehole reconstructions show greater warming in North America and Asia compared to Europe (Huang et al., 2000). A recent temperature reconstruction for the past 2000 years, from an alpine speleothem oxygen isotope record suggests that there were periods in that region of Europe when temperatures were more comparable (or even warmer) than today (Mangini et al., 2005), although this result is specifically dependent on the continuity of contribution of the same types of weather systems throughout this record.”
“The difference in amplitude of temperature variations within proxy reconstructions has been of recent concern (von Storch et al., 2004; Esper et al., 2005), with a number of factors in methodological procedures highlighted as being responsible for anomalies in the magnitude of temperature change within palaeoclimate records. The speleothem reconstruction displayed here has not been specifically manipulated to retain low-
frequency temperature variations or to preserve higher amplitude behaviour, but it does show a greater amplitude change over the past 500 years than that displayed by reconstructions that have been derived using “traditional’ regression methods. Thus, the larger temperature range present within the speleothem reconstruction over the last 500 years provides further support for the capture of a more slowly evolving climate signal within the speleothem records.”
Paul, the trouble with the time period of this study is that the 500-year period is not really in dispute. Why wouldn’t they have tried to cover the MWP? MAngini has a speleothem with a warm MWP?
Steve, I asked Claire Smith about the MWP:
Thanks for your interest. The results are already published on the
internet as a short communication which I’ve attached for your interest.
The actual data are available from the World data center
http://www.ncdc.noaa.gov/paleo/data.html along with many other
previously published well known climate reconstructions.
Unfortunately the length of our reconstruction is limited due to
available data. Yearly growth of stalagmites is very small (~10s of
microns) so actually extracting the annual changes in growth rate is
quite a tedious and time consuming task, hence the lack of datasets and
their short length. One aim of the article is to encourage the
production of more datasets from stalagmites.
One possibility for extending the reconstruction further back in time is
to use isotope or other geochemical information from stalagmites as
well. There a number of recent climate-isotope studies from stalagmites
which do extend much further back in time (glacial-interglacial time
scales) so the potential is there to further develop the work. I am
nearing the end of my PhD so I’m not sure how much I will achieve before
then as this is just one aspect of stalagmite-paleoclimate work I have
been investigating, however my supervisor, Ian Fairchild, has many
collaborators who are doing similar work and I know that they have
recently retrieved some very interesting data from a stalagmite from
Gibraltor which I think covers the Holocene period so would contain the
periods of interest.
I hope that helps to explain a bit of the background to paleoclimatology
from stalagmites. Please let me know if you require any further
information
Thanks
Claire
I also asked Ian Fairchild:
Dear Paul,
Thanks for your message. The work on Gibraltar is being led by my
colleague Dave Mattey from Royal Holloway who has obtained a beautiful
record – but only of the last 50 years! New material has been collected
to try to reach back to the last 1000 years. In this region though,
wetter/drier is more likely to be the key result rather than
warmer/colder. The consensus more generally at the moment is that the
Medieval Warm Period can be distinguished, but was not as warm as the
late 20th century.
Best wishes,
Ian
Do you know why they didn’t include Mangini’s stalagmite in their compilation – which has warm MWP?
Steve,
This might answer your question:
Click to access IJFInaugural.pdf
or maybe because they were only looking at the past 500 years, plus they say of Mangini – “this result is specifically dependent on the continuity of contribution of the same types of weather systems throughout this record.”
RE: 91
Willis,
It should be noted that Squires Gate and Ringway are airports, so likely have UHL issues, especially Ringway (the old name for Manchester Airport) which has expanded considerably since 1974.
Mike
Why is the last comment relevant to the exclusion of Mangini? The same would be true for the sites that they did use? The use of a Frisia series rather than Mangini seems completely arbitrary – and all too typical of the proxy field.
Another point which may or may not make a difference to results. It looks to me like the Smith collaiton as archived here http://www1.ncdc.noaa.gov/pub/data/paleo/speleothem/nhtemp-smith2006.txt
has incorrectly collated the Frisia information as archived here – if the archived version hasn’t been changed:
ftp://ftp.ncdc.noaa.gov/pub/data/paleo/speleothem/europe/italy/ernesto2003.txt
The series are the same, but the collated dates are different.
I’m not saying their comment on Mangini is a relevant excuse for not including it – I just posted what they said.
I don’t know what happened with the data archive, but both your plots are obviously of the Italian data.
This news article on my ISP site (in New Zealand), by Frank Hansen now claims the Earth is now at its warmest for the last million years
http://xtramsn.co.nz/technology/0,,13440-6352563,00.html
Neil
1. NO it does NOT. It says:
But what’s a degree or two?
2. The author was James Hansen.
I would say things are really heating up. I don’t like the cold winters here in New Brunswick but I love the summers and don’t really want them getting any hotter either.
This year we had record runs of Atlantic salmon returning to our rivers. Even to rivers they haven’t been back to for decades. It would really suck if global warming were to mess it all up.
James Mann,
It sounds like you believe CO2 emissions could result in warmer weather that could cause a problem with your hobby.
How many times more would you be willing to pay for gas, electricity and home heat so that this might not happen. Two times the current rates? Three? Four? Five?